DeepFuse: A multi-rater fusion and refinement network for
computing silver-standard annotations

J 2025

DeepFuse: A multi-rater fusion and refinement network for computing silver-standard annotations

AKBAS, Cem Emre, Vladimír ULMAN, Martin MAŠKA a Michal KOZUBEK

Základní údaje

Originální název

DeepFuse: A multi-rater fusion and refinement network for computing silver-standard annotations

Autoři

AKBAS, Cem Emre, Vladimír ULMAN, Martin MAŠKA a Michal KOZUBEK

Vydání

Computers in Biology and Medicine, Elsevier, 2025, 0010-4825

Další údaje

Jazyk

angličtina

Typ výsledku

Článek v odborném periodiku

Obor

10201 Computer sciences, information science, bioinformatics

Stát vydavatele

Nizozemské království

Utajení

není předmětem státního či obchodního tajemství

Odkazy

URL

Impakt faktor

Impact factor: 7.000 v roce 2023

Organizační jednotka

Fakulta informatiky

DOI

http://dx.doi.org/10.1016/j.compbiomed.2025.110186

Klíčová slova anglicky

deep learning; convolutional neural networks; label fusion; silver-standard annotations; computer vision

Štítky

cbia-web

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 28. 4. 2025 17:20, prof. RNDr. Michal Kozubek, Ph.D.

Anotace

V originále

Achieving a reliable and accurate biomedical image segmentation is a long-standing problem. In order to train or adapt the segmentation methods or measure their performance, reference segmentation masks are required. Usually gold-standard annotations, i.e. human-origin reference annotations, are used as reference although they are very hard to obtain. The increasing size of the acquired image data, large dimensionality such as 3D or 3D + time, limited human expert time, and annotator variability, typically result in sparsely annotated gold-standard datasets. Reliable silver-standard annotations, i.e. computer-origin reference annotations, are needed to provide dense segmentation annotations by fusing multiple computer-origin segmentation results. The produced dense silver-standard annotations can then be either used as reference annotations directly, or converted into gold-standard ones with much lighter manual curation, which saves experts’ time significantly. We propose a novel full-resolution multi-rater fusion convolutional neural network (CNN) architecture for biomedical image segmentation masks, called DeepFuse, which lacks any down-sampling layers. Staying everywhere at the full resolution enables DeepFuse to fully benefit from the enormous feature extraction capabilities of CNNs. DeepFuse outperforms the popular and commonly used fusion methods, STAPLE, SIMPLE and other majority-voting-based approaches with statistical significance on a wide range of benchmark datasets as demonstrated on examples of a challenging task of 2D and 3D cell and cell nuclei instance segmentation for a wide range of microscopy modalities, magnifications, cell shapes and densities. A remarkable feature of the proposed method is that it can apply specialized post-processing to the segmentation masks of each rater separately and recover under-segmented object parts during the refinement phase even if the majority of inputs vote otherwise. Thus, DeepFuse takes a big step towards obtaining fast and reliable computer-origin segmentation annotations for biomedical images.

Návaznosti

EF18_046/0016045, projekt VaV

Název: Modernizace národní infrastruktury pro biologické a medicínské zobrazování Czech-BioImaging

GA21-20374S, projekt VaV

Název: Segmentace a sledování buněk se složitým tvarem

Investor: Grantová agentura ČR, Segmentace a sledování buněk se složitým tvarem

LM2023050, projekt VaV

Název: Národní infrastruktura pro biologické a medicínské zobrazování

Investor: Ministerstvo školství, mládeže a tělovýchovy ČR, Czech BioImaging: Národní výzkumná infrastruktura pro biologické a medicínské zobrazování

90140, velká výzkumná infrastruktura

Název: e-INFRA CZ

Citovat

AKBAS, Cem Emre, Vladimír ULMAN, Martin MAŠKA a Michal KOZUBEK. DeepFuse: A multi-rater fusion and refinement network for computing silver-standard annotations. Computers in Biology and Medicine. Elsevier, 2025, roč. 192, Part A, s. 1-11. ISSN 0010-4825. Dostupné z: https://dx.doi.org/10.1016/j.compbiomed.2025.110186.

@article{2493619,
   author = {Akbas, Cem Emre and Ulman, Vladimír and Maška, Martin and Kozubek, Michal},
   article_number = {Part A},
   doi = {http://dx.doi.org/10.1016/j.compbiomed.2025.110186},
   keywords = {deep learning; convolutional neural networks; label fusion; silver-standard annotations; computer vision},
   language = {eng},
   issn = {0010-4825},
   journal = {Computers in Biology and Medicine},
   title = {DeepFuse: A multi-rater fusion and refinement network for computing silver-standard annotations},
   url = {https://www.sciencedirect.com/science/article/abs/pii/S0010482525005372},
   volume = {192},
   year = {2025}
}

TY  - JOUR
ID  - 2493619
AU  - Akbas, Cem Emre - Ulman, Vladimír - Maška, Martin - Kozubek, Michal
PY  - 2025
TI  - DeepFuse: A multi-rater fusion and refinement network for computing silver-standard annotations
JF  - Computers in Biology and Medicine
VL  - 192
IS  - Part A
SP  - 1-11
EP  - 1-11
PB  - Elsevier
SN  - 00104825
KW  - deep learning
KW  - convolutional neural networks
KW  - label fusion
KW  - silver-standard annotations
KW  - computer vision
UR  - https://www.sciencedirect.com/science/article/abs/pii/S0010482525005372
N2  - Achieving a reliable and accurate biomedical image segmentation is a long-standing problem. In order to train or adapt the segmentation methods or measure their performance, reference segmentation masks are required. Usually gold-standard annotations, i.e. human-origin reference annotations, are used as reference although they are very hard to obtain. The increasing size of the acquired image data, large dimensionality such as 3D or 3D + time, limited human expert time, and annotator variability, typically result in sparsely annotated gold-standard datasets. Reliable silver-standard annotations, i.e. computer-origin reference annotations, are needed to provide dense segmentation annotations by fusing multiple computer-origin segmentation results. The produced dense silver-standard annotations can then be either used as reference annotations directly, or converted into gold-standard ones with much lighter manual curation, which saves experts’ time significantly. We propose a novel full-resolution multi-rater fusion convolutional neural network (CNN) architecture for biomedical image segmentation masks, called DeepFuse, which lacks any down-sampling layers. Staying everywhere at the full resolution enables DeepFuse to fully benefit from the enormous feature extraction capabilities of CNNs. DeepFuse outperforms the popular and commonly used fusion methods, STAPLE, SIMPLE and other majority-voting-based approaches with statistical significance on a wide range of benchmark datasets as demonstrated on examples of a challenging task of 2D and 3D cell and cell nuclei instance segmentation for a wide range of microscopy modalities, magnifications, cell shapes and densities. A remarkable feature of the proposed method is that it can apply specialized post-processing to the segmentation masks of each rater separately and recover under-segmented object parts during the refinement phase even if the majority of inputs vote otherwise. Thus, DeepFuse takes a big step towards obtaining fast and reliable computer-origin segmentation annotations for biomedical images.
ER  -

AKBAS, Cem Emre, Vladimír ULMAN, Martin MAŠKA a Michal KOZUBEK. DeepFuse: A multi-rater fusion and refinement network for computing silver-standard annotations. \textit{Computers in Biology and Medicine}. Elsevier, 2025, roč.~192, Part A, s.~1-11. ISSN~0010-4825. Dostupné z: https://dx.doi.org/10.1016/j.compbiomed.2025.110186.

Přehled o publikaci